* don't ignore pfx files, we have few of them in these repo (certificate files used by networking benchmarks)
* change folder structure: corefx -> libraries, coreclr -> runtime
* change categories names
* update prerequisites docs
* update benchmarkdotnet docs
* update the microbenchmark design guidelines docs
* update the Microbenchmarks readme page docs
* update profiling workflow docs
* update benchmarking workflow docs
* remove old benchmarking workflow doc
* update README.md to point to the new docs
* Apply suggestions from code review
Co-Authored-By: Bill Wert <billwert@microsoft.com>
* revert accidental TFM removal
* dotnet runtime -> dotnet/runtime
* update the categories description
* change Categories names for real
Co-authored-by: Bill Wert <billwert@microsoft.com>
* add 3 more channels to microbenchmark windows job
* change framework to be required in benchmarks_ci.py
* move setting up pipeline from channel_map to ci_setup
* remove extra build status from README
* make ci_setup.py accept only one channel
* 3.0 should be downloaded from 3.0 channel, not 3.1
* add 3.1 to build matrix
* update build status page
* get the hash for 3.1 the same way we do for 3.0
* run ML.NET and Roslyn correctness jobs for 3.1, not 3.0
* run private non-scheduled jobs for 3.1, not 3.0
* run scheduled jobs for 5.0, not 3.0
* add 3.1 for the scheduled private jobs
* fix the build status matrix
* reflect Ubuntu 16 -> 18 change
* Ubuntu -> ubuntu
* RS4 -> RS5
* test x86 for PRs for .NET Core 5.0 only
* add Roslyn Ubuntu legs
* update BenchmarkDotNet to latest version to take advantage of improvements
* use FullNameProvider.GetBenchmarkName without reflection since it's public now
* update BDN to version that supports .NET Core 5.0
* add 5.0 to the matrix to make sure that this PR really fixes to 5.0 problem
* Add private jobs for microbenchmarks for netcoreapp5.0
* Use partition filter to split jobs into workitems
This change does the following:
* Uses helix.proj to split the work into 5 different work items for each
leg
* Removes the --allCategories parameter in favor of --filter * to
better use the partition filter
* Moves all of the windows jobs that use .Net Core to the base Windows
queue, since they do not require .Net Framework
* remove unnecessary code
* Update README with new links
* remove todo comment
* Stop supporting netcoreapp2.0
2.0 has gone out of support for .Net Core, so we should stop supporting
it in our performance testing framework. In particular, when you pull
down dotnet with -Channel 2.0, you are actually pulling down a dotnet at
version 2.1.200. There is an issue with this version where, if you run
on a machine where another run has executed nuget in the last 30 minutes,
you will get various errors of the "error NU3004: The package is not
signed" variety, which is causing a large number of our builds to fail.
Additionally, since netcoreapp2.0 is officially no longer supported, we
are not even running functional testing on the release/2.0 branches out
of coreclr or corefx. Therefore, we should not be running functional
testing of our performance tests for the netcoreapp2.0 bits.
This change removes netcoreapp2.0 support from the performance repo and
updates docs to reflect this.
* Remove 2.0 from Microbenchmarks.csproj
* Remove NETCOREAPP2_0 macro
* add path to the csproj with benchmarks to the JOB definition
* make categories a benchmark job parameter
* introduce ML.NET CI legs
* make the CI jobs IDs unique
* add ML.NET benchmarks to status page
* the BuildConfig needs to be unique to avoid logs overwriting in mc.dot.net
* move the explanation why we have a single repo to the main README.md
* move common.props prerequisites to root\build and reuse it for all working benchmark projects
* add comments to our common settings to explain why
* Create a single file that describes Prerequisites
* create a single doc page that describes BenchmarkDotNet
* simplify the micro benchmarks docs
* fix typos, grammar and broken links
* mention statistical test in the microbenchmarks README
* code review fixes
Adam Sitnik
Bill Wert
Olivia Chen
Adam Sitnik
Michelle McDaniel
José Rivero
Alaa Masoud